On the prevalence of Condorcet’s paradox

Abstract

The Condorcet paradox has been a significant focus of investigation since Duncan Black and Kenneth Arrow rediscovered its importance for economic and political theory. Recent research on this phenomenon has oscillated between simulation studies, probability calculations based on hypothetical voter preferences, and empirical analyses of single election studies. This paper presents a comprehensive evaluation of 253 electoral polls conducted across 59 countries. Our findings demonstrate that the Condorcet paradox has virtually no empirical relevance: we find no robust evidence of cyclical majorities in any of the 253 elections. This result remains robust after statistical inference testing. Furthermore, this study provides insights into which parties are particularly likely to emerge as Condorcet winners and explores how these Condorcet winners assert themselves after elections.